https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7971

Bill Cole <billc...@apache.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|3.4.4                       |SVN Trunk (Latest Devel
                   |                            |Version)
                 CC|                            |billc...@apache.org
                 OS|Linux                       |All
           Hardware|PC                          |All
           Priority|P2                          |P3
           Severity|normal                      |minor

--- Comment #1 from Bill Cole <billc...@apache.org> ---
DOS_RCVD_IP_TWICE_B has not changed since 2008. See
https://svn.apache.org/viewvc/spamassassin/trunk/rulesrc/sandbox/dos/70_other.cf?r1=627944&r2=627945&;

That rule depends on specific mail routing details and configuration of local
parameters like trusted_networks and internal_networks, so it is impossible to
analyze what is causing you to see a lot of hits on that without full sample
emails. I do not see a large number of hits on this rule in the systems I work
with. 

However, I do see *some* hits that are on definite ham, resulting from local
mail submission on a public address. That's not common but it is also not
"wrong" and in this specific case there's a solid reason for it.  

Looking at RuleQA I see that the rule is fairly reliable and hits a large
amount of spam, but it also has substantial hits on ham at most reporting sites
(as much as 2.5% of all ham!) and hits only ham at a few. 

I've limited the score to 2.0 in revision 1899866. I am very reluctant to
modify the rule to reduce its hits on ham based solely on the idiosyncratic
examples that I have in hand from 1 source. If you have matching non-spam
samples that you can share, please attach them to this ticket so that we can
(maybe) refer to them and modify the rule to avoid problems.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to